Stage 4 · Mastery I · Quantitative Edge
₹39,999 · 5 volumes · 12 Python notebooks · 8-week capstone · Lifetime access
Stage 4: turn the documented playbook into code, with statistical integrity.
Stage 3's playbook is judgment-driven application of documented rules. Stage 4 turns those rules into Python — backtested with statistical integrity, regime-conditioned, factor-decomposed, and stress-tested via Monte Carlo. The transition from discretionary to systematic isn't about removing judgment; it's about isolating the parts that can be coded from the parts that can't.
What you can do after this stage
- Build Python backtests with FY26 Indian transaction-cost stack
- Apply CPCV and Deflated Sharpe Ratio to every strategy before deployment
- Translate discretionary setups into statistically validated quantitative systems
The five Stage 4 volumes
The minimum viable Python toolkit
Pandas for time-series. Numpy for vector math. Matplotlib for visual diagnostics. Scikit-learn for regression and classification. Statsmodels for cointegration and stationarity tests. The volume assumes no prior Python; by the end of it, you can read OHLCV data into a DataFrame, compute technical indicators from scratch, and visualise a strategy's equity curve. Includes 12 Jupyter notebooks — each ~80 lines — that you run and modify directly.
Why most retail backtests are useless
Look-ahead bias. Survivorship bias. Data-snooping bias. In-sample vs out-of-sample distinction. Walk-forward analysis. The four backtest results that look good but mean nothing. Includes the Sharpe ratio's blind spots, why max drawdown matters more than CAGR for retail, and the Lopez de Prado sequential-bootstrap test for assessing whether observed edge is statistically distinguishable from luck.
Model the time-series, then decompose returns into components you understand
The time-series toolkit first: stationarity and the ARIMA family, volatility clustering with GARCH, and cointegration for pairs. Then the cross-sectional layer: why a strategy's return is not one number but a decomposition — market beta + factor exposures + idiosyncratic alpha. The four canonical factors (size, value, momentum, quality) and how to compute factor exposures from a backtest, including regime-conditional factor analysis. Includes the Indian-market specifics — Nifty 50 has materially different factor structure from S&P 500.
Treat the model as cross-sectional ranking, validated against overfitting
Machine learning framed the way it actually survives in markets: a monthly cross-sectional ranking problem, not a price-prediction oracle. Feature engineering from price, volume, and fundamentals. Tree-based models and where they break. The validation discipline that separates a real model from a curve-fit — purged and embargoed k-fold cross-validation, leakage detection, and reading feature importance with SHAP. Every model is a Python notebook you run and modify on your own data.
Prove the edge survives before live capital touches it, then close the systematic build
The capstone gate: Probability of Backtest Overfitting, the Deflated Sharpe Ratio, and Combinatorial Purged Cross-Validation applied to your own strategy. Monte Carlo robustness as part of the same pass — trade-sequence resampling for the drawdown distribution, the expected vs worst-case gap, and risk-of-ruin conditional on percent-risk-per-trade. Then the systematic-build close: take one stable Stage 3 setup through the full pipeline, paper-trade for two weeks, and submit code + journal for grading. Pass = automatic Stage 5 discount code. Fail = re-grade with detailed feedback; lifetime access means no penalty for re-takes.
What Stage 4 is not
- Not a Python bootcamp. It's a systematic-trading curriculum that uses Python as the implementation language. We assume zero Python; we don't aim to produce software engineers.
- Not a full algorithmic trading framework. Stage 4 builds the analytical foundation; Stage 5 (Systems Architect) builds the execution architecture.
- Not a substitute for trading judgment. The systematic translation isolates the codeable parts and preserves the parts that aren't. Most edge in retail trading lives in the regime-classification and risk-overlay layers, which remain human-led at Stage 4.
Who should buy Stage 4 right now
- You completed Stage 3 capstone. Your 25-setup playbook has been running for at least 4 weeks. You're ready to translate the most stable setup into code.
- You want backtesting integrity. Most retail backtest claims you've seen are statistically wrong. Stage 4 fixes that gap.
- You want to understand why a strategy works, not just that it works. The factor decomposition + Monte Carlo workflow is the institutional answer.
- You have any aspiration toward AIF Cat III, fund management, or a SEBI-RA practice. Stage 4 is the analytical floor for those careers.
Who should NOT buy Stage 4 yet
- You haven't passed Stage 3 capstone. The systematic translation requires a documented setup to translate — without that, Stage 4 has no inputs.
- You want a black-box trading system delivered. Stage 4 builds your system using your inputs; we don't sell pre-made strategies.
- You're allergic to code. The 12 notebooks are not optional. We try to make Python accessible; we can't make it unnecessary.
Enrol in Stage 4
₹39,999 · 5 volumes · 12 Python notebooks · 8-week capstone · Lifetime access
Enrol Stage 4 — ₹39,999Bharath Shiksha is an educational publisher. We do not provide investment advice. The curriculum uses anonymised historical examples with at least 30-day data lag; no specific securities are named for buy/sell/hold; no performance claims, return projections, or accuracy statistics are made. Trading involves substantial risk of capital loss.